ARTICLE IN PRESS

Nuclear Instruments and Methods in Physics Research A 512 (2003) 546–552

Neutron detection with a silicon PIN photodiode and 6 LiF converter M. Voytcheva,*, M.P. In˜igueza, R. Me! ndeza, A. Man˜anesb, L.R. Rodr!ıguezc, R. Barquerod a

! ! Dpto de F!ısica Teorica y Atomica, Nuclear y Molecular, Universidad de Valladolid, 47011 Valladolid, Spain b Dpto de F!ısica Moderna, Universidad de Cantabria, 39005 Santander, Spain c Dpto de F!ısica, Universidad de Burgos, Avda. Cantabria s/n., 09006 Burgos, Spain d ! Radiologica ! Servicio de Proteccion del Hospital R!ıo Hortega, 47011 Valladolid, Spain Received 18 November 2002; received in revised form 11 May 2003; accepted 12 May 2003

Abstract PIN silicon photodiodes can be used successfully for thermal neutron measurements with 6LiF converters. Tritons created in the converter are detected by the diode with high efficiency and energy resolution. An optimal separation for the triton peak from the alpha particle one is obtained with a converter about 2 mm thick. The background in these peaks due to gamma radiation or products from fast neutron reactions inside the detector is less than 1%. Experimental studies with different converter thicknesses and at different distances between the Am–Be source and the detector show good agreement with Monte Carlo simulations and with TLD measurements. The detector sensitivity is 3.270.5  10 3 counts in the triton peak per unit fluence of thermal neutrons. r 2003 Elsevier B.V. All rights reserved. PACS: 29.30.Hs Keywords: Neutron; Silicon; Photodiode; 6LiF; Converter; Triton

1. Introduction Commercial PIN silicon photodiodes are used to detect visible light but they can also be used to detect radiation and particularly alpha particles [1–4]. These photodiodes generally have low cost, very good quantum efficiency and energy resolution. They can work with a few volts bias voltage and could have a dead entrance layer of several *Corresponding author. Address for correspondence: Bat. 516, CEA-Saclay, 91191 Gif sur Yvette, France. Tel.: ++33-169-08-46-81; fax: ++33-1-69-08-60-30. E-mail address: [email protected] (M. Voytchev).

tens of nanometers. The thin depletion layer (200– 500 mm) gives them a low sensitivity for gamma rays but it is thick enough for full collection of the electron–hole pairs generated by charged particles. Thus, for many types of alpha radiation measurements, the background counting rate (due to other types of radiation) in the energy region up to 1 MeV can be less than 1 count per 10 h [4]. On the other hand, photodiodes are sensitive to light and most importantly to environmental electromagnetic noise. The latter can be shielded with a metallic box connected to the ground of the associated electronics.

0168-9002/$ - see front matter r 2003 Elsevier B.V. All rights reserved. doi:10.1016/S0168-9002(03)02013-8

ARTICLE IN PRESS M. Voytchev et al. / Nuclear Instruments and Methods in Physics Research A 512 (2003) 546–552

For the detection of thermal or low energy neutrons, photodiodes can be utilized in combination with 6LiF (used in this work) or 10B converters placed adjacently to the detector’s surface. Neutrons interact through exothermic reaction 6Li(n,a)3H with the converter and produce 2.05 MeV alpha particles and 2.73 MeV tritons (3H) which are easily detected by the photodiode. Boron converter could also be used although the capture reaction producing 480 keV gamma rays because the energy interval below about 500 keV is generally not used for radiation detection (this will be discussed in the Section 3). However, the tritons and alpha particles from the 6 Li reaction have higher energy than the 1.47 MeV alpha particles from the 10B reaction and are finally more suitable for detection with these photodiodes. This paper will present different measurements with a Hamamatsu photodiode coupled with 6LiF converters (95% 6Li enrichment) of different thicknesses, irradiated with a moderated Am–Be neutron source. Monte Carlo simulations were performed to calculate the neutron fluence rate at the irradiation positions and to determine the spectrum of the energy deposited in the detector.

2. Materials and methods A commercial windowless silicon photodiode Hamamatsu S3590-02, p–i–n junction type, was used as detector. The active area of the diode was 1 cm2 (1 cm  1 cm), the depth of its depletion zone—200 mm. The measuring system included also a Hamamatsu H4083 charge preamplifier and a conventional multichannel analyzer including an amplifier. The energy calibration of the spectrometer was performed by using 241Am standard solid source as well as with 218Po (6.0 MeV) and 214Po (7.7 MeV) alpha peaks obtained during radon measurements [4]. Three 6LiF converters with different thicknesses were prepared for use with the detector. The 300 mm one was moulded by compressing 6LiF powder, subsequently heated at 600 C for 2 h for solidification. Two thinner converters, respectively, with 2 and 0.6 mm thicknesses, were

547

fabricated by atomic deposition. 6LiF was evaporated on an aluminum support (thickness 1 mm, surface 1.4  1.4 cm2) and sealed with a silver layer 10 nm thick. A moderated 241Am–Be neutron source that supplies about 104 thermal neutrons per second to the detector was used for irradiation. The yield of the source was of the order of 106 neutrons s 1 and a goal of this study was to determine it exactly. The cylindrical source with a diameter and height of 3.5 cm (iron shielding of 0.4 cm included), was put inside a 60 cm  60 cm  75 cm paraffin parallelepiped covered by three, 2 mm thick layers of cadmium, plastic and iron (Fig. 1). The source is located 45 cm from the upper side of the paraffin container and 30 cm from the other border surfaces. The detector can be placed at different distances from the source through a cylindrical air channel (diameter 4.5 cm) where the space in the channel between the source and the detector can be filled with paraffin. The position of the detector in all measurements is 1 mm from the source except the cases where different positions are expressly mentioned. Measurements are also carried out with thermoluminescent LiF detectors TLD600 and TLD700. The procedure of thermal neutron fluence rate determination is explained in Ref. [5]. Connection towards the amplifier and multichannel analyzer 4.5 cm large cylindrical air channel that can be filled with paraffin

45 cm

Box with the diode, converter and preamplifier

75 cm

2 mm iron 2 mm plastic 2 mm Cd paraffin

30 cm Am-Be source

60 cm

Fig. 1. Schema of the used Am–Be neutron source with its shielding and the detector position.

ARTICLE IN PRESS M. Voytchev et al. / Nuclear Instruments and Methods in Physics Research A 512 (2003) 546–552

548

Monte Carlo simulations of the interaction of incident neutrons and of the detection of tritons and alpha particles were performed respectively by the MCNP4C code [6] and by an original software program ALPHASIM [7] using data from the ion transport code TRIM [8,9] (the most recent version of this code SRIM2003 is available at http://www.srim.org).

3. Results and discussion Fig. 2 shows the neutron spectrum simulated with the MCNP code at the detector position as well as the spectrum of the Am–Be source [10]. The simulation geometry reproduces strictly the experimental one. Due to the alpha-beryllium reactions and to the paraffin moderation, neutrons that will irradiate the converter and the detector will have energies from the thermal region up to about 10 MeV. Neutrons interact with the converter and the silicon photodiode, including doping materials and impurities. Before manufacturing 6LiF converters, some simulations were carried out in order to determine their optimal thickness. A spectrum with two energy peaks of about 2 MeV (for alpha particles) and about 2.7 MeV (for tritons) was expected (see Section 1). The separation of just one of those peaks from the rest of the spectrum would be

Neutron fluence rate (cm-2 s-1)

10-2 Am-Be source spectrum x 10-8

10-3 MCNP simulated neutron spectrum at detector position around the used Am-Be source

10-4

10-5 10-8 10-7 10-6 10-5 10-4 10-3 10-2 10-1

1

101

Energy (MeV)

Fig. 2. Initial Am–Be and MCNP simulated at detector position neutron spectra. Each point of the figure represents the total fluence rate in the energy interval between the previous and the current energy.

enough for the determination of the number of thermal neutrons interacted in the converter because this number is proportional to the peak total area. Since the background in the considered cases is rather low (it will be seen in the following), the principal separation is, therefore, between the two peaks. Since tritons have a longer range in LiF (about 30 mm) than alpha particles (about 6 mm) their peak will have a better resolution, higher energy and its consideration is easier. A thinner converter would produce a triton peak well distinguished from the alpha one but with a lower counting rate. To optimize the thickness of the converter we used ALPHASIM program that was initially developed for simulation of the detection of radon and its progeny. This was with the same photodiode and it gave good results [7]. The program assumes that monoenergetic alpha particles and tritons are created homogenously in the volume of the converter. Then, the program simulates the particle emission from each of small elementary volumes on which is divided the converter, choosing randomly the direction of emission. Taking into account the energy losses of the charged particles in the converter (using data from the TRIM code), the geometrical conditions and the minimal detection angle of the photodiode, the program determines if the particle will be detected and with which energy. Simulating the obtained spectrum for different converter thicknesses, we saw that the maximal thickness where both the peaks are still well separated is around 2 mm (Fig. 3, part A). Fig. 4 shows the simulated spectrum that is obtained in 5 min by a photodiode with a 6LiF converter of 2 mm thickness located at 1 mm distance from the diode if 100 alphas and tritons are created homogenously, per second, in the converter volume. With the same program we also estimated the reduction of the counting rate in the triton peak when the converter thickness decreases (Fig. 3, part B). It can bee seen that for the converter of 2 mm, the number of counted tritons would be about 30% of the maximal number obtained for a thick enough converter (10 mm or more). Let us consider first the experimental irradiation of a bare photodiode that is without any converter, inside the neutron field. Fig. 5 shows

ARTICLE IN PRESS

(B) Counting rate vs. maximal counting rate for a thick enough converter (10µm and more) in the triton peak

200

(A) Overlapped area from the triton peak in the alpha one

5

10 15 20 25 Converter thickness (µm)

30

35

Fig. 3. Overlapped area from the triton peak in the alpha one and counting rate vs. maximal counting rate in the triton peak as a function of the converter thickness.

300 250 200 Counts

150 100 50

0

150 100 50 0

549

250

100 90 80 70 60 50 40 30 20 10 0

Counts

Percentage (%)

M. Voytchev et al. / Nuclear Instruments and Methods in Physics Research A 512 (2003) 546–552

0

500

1000 1500 2000 Energy (keV)

2500

3000

Fig. 4. Simulated spectrum that would be obtained for 5 min by a photodiode with a 6LiF converter of 2 mm thickness and 1.4 cm  1.4 cm size situated at 1 mm distance from the diode, if 100 alphas and tritons are created homogeneously per second, in the converter volume. Counts below 500 keV are discriminated.

the spectrum of the energy deposited in the bare diode after 12 h of irradiation. The peak around 1.8–2 MeV is due to the reaction 10B(n,a)7Li where neutrons are captured by boron dopants existing in the photodiode. The reactions of fast neutrons with silicon, like 28Si(n,p)28Al and 28Si(n,a)28Al contribute to that peak as well to the tail on the right part of the spectrum. The lower energy part of any spectrum acquired with this photodiode is generally due to electronic noise in the detector. With appropriate electromagnetic shielding and adjusting of the amplification, this noise can be

0 0 1000 2000 30004000 5000 6000 7000 8000 9000 Energy (keV) Fig. 5. Energy distribution of the charged particles produced after 12 h of irradiation of a bare photodiode with neutrons with energies from thermal region up to 10 MeV.

decreased and shifted to lower energies, though not less than about 0.5 MeV. As it can be seen also in Ref. [4], the energy interval 0–0.5 MeV is not suitable for radiation detection by these photodiodes and it is discriminated in the spectra discussed in the following. However, in the spectrum of Fig. 5, one can see an important contribution in the part up to about 1.2 MeV. This is due to the gamma rays of the activation process that happens in the aluminum box used around the detector. It is important to note that the spectrum of Fig. 5 represents the background counting in the measurements with converters. It gives only about 1% of the total counting in the spectra with converters shown in the following (Figs. 6–8) and because of that its other particularities will not be considered in this study. Further details of measurements with a bare photodiode can be found in Ref. [11]. Figs. 6–8 show, respectively, experimental spectra obtained with the photodiode and three different converters (with a thickness, respectively, 300, 2 and 0.6 mm) placed at 1 mm from the Am– Be neutron source (Fig. 1). The dependence of the spectrum shape on the converter thickness can be observed and it can be seen how both peaks become more separated with the decrease of the converter thickness. Experimental results confirm simulation ones that an optimal converter thickness is approximately 1–2 mm. Comparing the

ARTICLE IN PRESS M. Voytchev et al. / Nuclear Instruments and Methods in Physics Research A 512 (2003) 546–552

550 1000

120

900

100

800

80

600

Counts

Counts

700 500 400

60 40

300 200

20

100 0 0

500

1000 1500 2000 Energy (keV)

2500

3000

Fig. 6. Energy distribution of the charged particles produced after 30 min of irradiation of a photodiode +300 mm thick 6LiF converter with neutrons with energies from thermal region up to 10 MeV. The counts in the range 3–8 MeV are negligible.

0 0

500

1000 1500 2000 Energy (keV)

2500

3000

Fig. 8. Energy distribution of the charged particles produced after 30 min of irradiation of a photodiode +0.6 mm thick 6LiF converter with neutrons with energies from thermal region up to 10 MeV. The counts in the range 3–8 MeV are negligible.

400 350

Counts

300 250 200 150 100 50 0 0

500

1000 1500 2000 Energy (keV)

2500

3000

Fig. 7. Energy distribution of the charged particles produced after 30 min of irradiation of a photodiode +2 mm thick 6LiF converter with neutrons with energies from thermal region up to 10 MeV. The counts in the range 3–8 MeV are negligible.

counting rates with those of Fig. 5, one can also see that background counting from gamma rays or other particles created in the diode through fast neutron reactions, is really negligible. Thus, the number of created tritons (and therefore the number of interacted thermal neutrons) can be accurately determined. The shape of the experimental spectrum from Fig. 7 can be compared with the shape of simulated one from Fig. 4 since both are for the same converter thickness (2 mm) and only a scaling coefficient on y-axis exists due to the different number of interacted neutrons.

Both shapes are very similar which is important for validation of the simulations. The differences are due to not taking into consideration the simulation of the background and the statistical processes in the photodiode during the detection of charged particles. Photodiode response was compared with that of calibrated thermoluminescent TLD-600 dosimeters [5] as well as with MCNP simulations. MCNP simulations show that more than 98% of all of the interactions in the converter producing charged particles come from thermal neutrons. This thermal neutron response of the photodiode coupled with 2 mm converter was compared in Fig. 9 with the thermal neutron response of TLD600, for different distances between the source and the detector where the cylindrical channel between them was filled with paraffin. MCNP simulations of the thermal neutron fluence rate at detector positions are also presented in that figure. Adjusting the curve of the photodiode counting rate (which is in counts min 1) with the TLD one for paraffin thicknesses of 5 cm and more, a calibration coefficient of 5.2715% was found for the photodiode response. This result shows that 310747 thermal neutrons cm 2 are needed to enter into the converter volume in order to produce one detected count in the 3H peak. Thus, the detector sensitivity is 3.270.5  10 3 counts in

ARTICLE IN PRESS

Thermal neutron fluence rate (cm-2 s-1)

M. Voytchev et al. / Nuclear Instruments and Methods in Physics Research A 512 (2003) 546–552 10000 8000

Diode triton peak counting rate (cts/min.) x callibation coefficient TLD

6000

MCNP simulation x yield of the source

4000 2000 0 0

5

10 15 20 Paraffine thickness (cm)

25

Fig. 9. Experimental (photodiode, TLD) and simulated (MCNP) thermal neutron response as a function of the paraffin thickness between the source and detector. Error bars include only statistical uncertainties.

551

precise measurements, especially for the case when the detector is very close to the neutron source, are envisaged. A good agreement is also found for the dependence of the counting efficiency with the thickness of the converter. The experimental 3H peak counting rates obtained for the 0.6 and 2 mm thick converters are, respectively, 7.9370.40 and 29.171.0 counts s 1, which shows a nearly linear dependence. First, this is due to the linear dependence of the number of produced tritons in the converter. And second, ALPHASIM simulating program shows that the same proportion (23%) of all emitted tritons in the volume of both converters will be detected by the photodiode.

4. Conclusion and perspectives the triton peak per unit fluence of thermal neutrons. Adjusting the curve of the MCNP fluence rate to the TLD one (because simulations were performed for 1 neutron s 1 emitted by the source) the exact yield of the Am–Be neutron source was determined—3.070.5  106 neutrons s 1. Obtained results for the different points from 5 to 23 cm are very close, which allows us to calibrate the photodiode response to the thermal neutron fluence rate. However, it has to be noted that when measurements are carried out at the nearest positions (less than 4–5 cm of paraffin thickness), exact calibration is not possible because the TLD results are for almost a point position, while the diode response is an integral over the whole converter and diode surface. The above experimental calibration can be easily compared with calculations. On one hand, from the simulations with ALPHASIM program, we obtain that 23% of whole generated in the converter volume, tritons are going to be detected by the photodiode and will constitute the 3H peak. On the other hand, considering thermal neutrons with energy less than 0.4 eV (see Fig. 2) and averaging a calculated cross-section of about 630 b for this thermal neutron region, a number of about 260 neutrons cm 2 per count in the triton peak can be easily calculated. The obtained theoretical value is close enough to the experimental one. More

PIN silicon photodiodes can be successfully used for thermal neutron measurements with 6LiF converters. Tritons created in the converter are very well detected by the diode. An optimal separation for the triton peak from the alpha one is obtained with a converter about 2 mm thick. In this case, the 3H peak counting rate would be about 30% of the same counting rate when a thick enough (more than 10 mm) converter is used. The background counting in this peak that comes from gamma radiation or products from fast neutron reactions inside the photodiode is less than 1%. Experimental studies with different converter thicknesses and with different paraffin layers between the source and the detector show a good correlation when compared with Monte Carlo simulations and with TLD measurements. About 23% of the whole number of generated tritons in the converter (1.4 cm  1.4 cm  2 mm) volume are detected by the photodiode and constitute the 3H peak. A detector sensitivity of 3.270.5  10 3 counts in triton peak per unit fluence of thermal neutrons is obtained. In order to make a neutron spectrometer sensitive to other energies (not only for thermal ones) we are working in two directions. On the qone hand, we are developing the use of a polyethylene converter producing protons, as it was already shown in Ref. [12], for the same type

ARTICLE IN PRESS 552

M. Voytchev et al. / Nuclear Instruments and Methods in Physics Research A 512 (2003) 546–552

of photodiodes. Our preliminary experiments also confirm that application. On the other hand, we are planning to put the small box with the diode, converter and preamplifier inside Bonner spheres of different diameters. In a previous work [4] a small and portable three channel alpha counter was developed for in situ radon measurements with this photodiode. The idea was that when knowing the spectrum that would be acquired a counter indicating only the total counting in a given energy interval is enough for a measurement. Since the detection process in both cases (radon measurements and thermal neutron detection through 6LiF converter) is practically the same and it is only a matter of changing the energy intervals, we are envisaging the use of the same counter for neutron measurements.

Acknowledgements This work has been partially supported by the Marie Curie Fellowship Program of the European Commission (ref. no. HPMD-GH-00-00032-01) and by the Consejo de Seguridad Nuclear de Espa n˜a. The authors also thank Mr. Michael Drummond for the editing of this manuscript.

References [1] H. Yamamoto, S. Hatakeyama, T. Norimura, T. Tsuchiya, Radioisotopes 33 (1984) 864. [2] P. Gooda, W. Gilboy, Nucl. Instr. and Meth. A 255 (1987) 222. [3] A. Chambaudet, D. Klein, M. Voytchev, Radiat. Meas. 28 (1997) 127. [4] M. Voytchev, D. Klein, A. Chambaudet, G. Georgiev, Health Phys. 80 (6) (2001) 590. [5] R. M!endez, M.P. In˜iguez, R. Barquero, A. Man˜anes, E. Gallego, A. Lorente, M. Voytchev, Rad. Prot. Dosim. 98 (2) (2002) 173. [6] J.F. Briesmeister (Ed.), Report LA-12625-M, Los Alamos National Laboratory, 1993. [7] M. Voytchev, D. Klein, K. Voytchev, Health Phys. 80 (6) (2001) 597. [8] J.P. Biersack, L.G. Haggmark, Nucl. Instr. and Meth. A 174 (1980) 257. [9] J.F. Ziegler, J.P. Biersack, U. Littmark, The Stopping and Range of Ions in Solids, Pergamon Press, New York, 1985. [10] ISO 8529:1989, Neutron reference radiations for calibrating neutron-measuring devices used for radiation protection purposes and for determining their response as a function of neutron energy. [11] S. Agosteo, A. Castoldi, L. Castellani, P. Colautti, G. D’Angelo, L. De Nardo, A. Favalli, I. Lippi, R. Martinelli, G. Tornielli, P. Zotto, Radiat. Prot. Dosim. 99 (1–4) (2002) 343. [12] S. Agosteo, L. Castellani, G. D’Angelo, I. Lippi, P. Zotto, Annual Report, Detectors and Instruments Development Division, Laboratoi Nazionali di Legnaro, Italy, 2001.